Research

Abstract: We study Argentina’s 2023-24 disinflation as a fiscal regime-shift episode. We use two electoral surprises that moved perceived fiscal-reform probabilities in opposite directions and show that inflation compensation repriced sharply around both events. The response is strongest at the medium term, where inflation forwards rise after news that reduced the perceived likelihood of fiscal reform and fall after news that increased it. We measure the fiscal content of the electoral shock using budget forecasts and narrative data from professional forecasters, which indicate a significant change in the policy regime. Expected future nominal policy rates also decline after Milei’s election, which is hard to reconcile with a standard monetary-tightening account. Instead, the joint fall in inflation compensation and expected nominal rates is consistent with a stabilization in which monetary policy becomes effective precisely because it is backed by fiscal adjustment. The subsequent realized decline in inflation from 25.5 percent to 4.3 percent within five months of the runoff validates the earlier market repricing, pointing to fiscal regime change as a central driver of Argentina’s stabilization.

Accessible here. Prepared as final project for Stanford's CS 229. 

Abstract: As large language models become central intermediaries in information acquisition, they may either aggregate dispersed beliefs into common answers or reproduce the heterogeneity of the users who query them. This project studies that distinction in the context of probabilistic forecasting. I examine how LLM forecasts respond to user-stated priors, public benchmark information, prompt wording, and model design. The analysis asks when LLM advice behaves like a common signal that compresses disagreement, and when it behaves like a personalized intermediary that tracks user beliefs or linguistic choices. The project contributes to a broader economics of AI-generated information by identifying the conditions under which algorithmic advice homogenizes beliefs, preserves disagreement, or reallocates informational influence across users.

Abstract: Prediction-market contracts look like state-contingent securities, but their usefulness for hedging depends on the actual menu, prices, and liquidity of listed claims. This paper studies that distinction using Kalshi CPI contracts, which pay when one-decimal reported inflation is above a threshold. I construct quote-consistent no-arbitrage intervals for inflation swaps, caps, floors, corridors, and tail payoffs, and use interval width as a model-free measure of hedgeability. Kalshi CPI prices contain useful information about inflation outcomes: the final Kalshi-implied YoY mean has a 0.968 correlation with realized YoY CPI, and calibration improves Brier scores for both YoY and MoM events. But the same contracts provide weak hedges in the full listed market. Median no-arbitrage widths are 9.69 percentage points for a YoY swap payoff and 4.95 for a YoY cap payoff. The actionable-looking market is much smaller: in baseline YoY observations, the median maximum quote age is 331 hours and the median number of quotes updated within 24 hours is zero. The results show why informative prediction markets need not be good hedging markets: sparse threshold claims, bid-ask frictions, stale quotes, and payoff mismatch leave many inflation exposures poorly spanned.

Undergraduate Honors Thesis (Binghamton) under supervision of Barry Jones.

Abstract: I extend Calvo and Velasco's (2022) model of monetary-fiscal coordination by incorporating explicit active (non-Ricardian) and passive (Ricardian) fiscal policy rules, following Leeper (1991). This extension provides richer bond price dynamics, revealing additional scenarios requiring policy coordination, while overturning others. Using analytical solutions to the model's differential equations, I demonstrate that monetary intervention is only necessary under non-Ricardian fiscal policy, while Ricardian fiscal policy is neutral. While this finding overturns some of Calvo and Velasco's results, it maintains the broader flavor about the interconnectedness of monetary and fiscal policy. The model provides insights into policy responses during the Covid-19 pandemic and offers a framework for analyzing future crisis responses.

Using the comprehensive treasury dataset from Hall, Payne, Sargent, and Szoke (2018), I examined how the Louisiana Purchase bonds— which expanded U.S. debt by approximately 20% (equivalent to 95% of annual government revenues)—set a precedent for U.S. sovereign debt as a global financial instrument. Unlike most pre-1920 U.S. bonds, which were illiquid and uniquely tailored to congressional authorizations, these bonds achieved unprecedented international marketability, trading in London, Paris, and Amsterdam. Facilitated by international merchant banks, this innovative structure enabled the U.S. to execute a massive financial expansion while maintaining fiscal credibility. This episode marked a key developmental stage in the evolution of U.S. Treasuries into the world’s predominant safe asset, illustrating how institutional credibility and strategic financial engineering can enable large-scale national expansion without destabilizing debt markets. This work was published as a Jupyter Notebook and can be accessed on Tom's website. 

Presented at 49th Eastern Economic Association Conference, 2022, November.